Cues used for distinguishing African American and European American voices

Abstract

Past studies have shown that listeners can distinguish most African American and European American voices, but how they do so is poorly understood. Three experiments were designed to investigate this problem. Recordings of African American and European American college students performing various reading tasks were used as the basis for stimuli in all three. In the first experiment, stimuli were subjected to monotonization, lowpass filtering at 660 Hz, and no modification. In the second, stimuli featuring certain ethnically diagnostic vowels and control stimuli were subjected to monotonization, conversion of vowels to schwa, or no modification. In the third, stimuli featuring diagnostic vowels and control stimuli were modified so that the intonation of paired African American and European American speakers was swapped. In all three experiments, African American and European American listeners in North Carolina and European American listeners in West Virginia identified the ethnicity of the speaker of each stimulus. Vowel quality emerged as the most consistent cue for identifications. However, listeners accessed other cues differently for male and female speakers. Breathiness was correlated with identifications of male speakers but not of female speakers. F0-related factors proved more important for female speakers than for male speakers. [Work supported by NSF.]